Industrial IoT Networks Shift Toward on-Module AI as Satellite Links Expand Global Coverage
Industry leaders are restructuring how they deploy connected infrastructure across manufacturing floors, logistics fleets, and remote sites. The shift comes as new on-module machine-learning SDKs and satellite IoT projections reshape the architecture that enterprises rely on for real-time decision-making. These changes are moving IoT from simple telemetry gathering to active, local intelligence that functions even where terrestrial signals fail.
On-Module ML Inference Reaches Cellular Hardware
Telit Cinterion has announced an edge AI SDK that runs machine-learning models directly on its cellular modules without external accelerators. Built on LiteRT and TensorFlow Lite, the software lets device manufacturers perform classification, anomaly detection, and time-series forecasting at the network edge. The SDK arrives in Q4 2026 and covers 4G, 5G RedCap, and high-performance 5G hardware. This development ensures that the modem is no longer just a communication pipe but a processing node in its own right.
That release matters because earlier cellular modules handled connectivity rather than computation. Factories using legacy deployments relied on gateways or cloud pipelines to analyze sensor data, which introduced latency and bandwidth costs. With on-module inference, telemetry streams can be filtered before transmission, reducing payload size while improving response times for time-critical alerts such as thermal runaway detection in battery systems or mechanical wear in assembly robots. This architectural change cuts down on the constant polling of cloud resources.
Telit Cinterion said the SDK gives developers C and Python APIs, model quantization tools, and OTA hooks so trained weights can be pushed to fleets without swapping hardware. Pilot customers in smart metering and cold-chain monitoring are testing vibration and temperature models on live lines ahead of general availability. Support spans several module families so existing certified designs can add inference with a firmware update rather than a board respin. In proof-of-concept testing, the company reported that inference consumed no more than 17% of CPU resources, which is vital for keeping hardware within thermal limits.
Satellite IoT Projections and the Global eSIM Expansion
Omdia projects satellite IoT connections to reach 197.6 million by 2035, driven by use cases that require tracking beyond terrestrial network coverage. Connected vehicles, offshore equipment, and agricultural sensor networks represent the largest addressable segments, according to analyst commentary published in August 2026. The shift points to a market that is moving beyond niche tracking toward high-volume applications where reach is the primary constraint rather than bandwidth.
At the same time, providers like Solve Networks are introducing global eSIM solutions to simplify the management of these far-flung assets. For organizations managing IoT devices across multiple countries, connectivity has traditionally required juggling separate SIMs and carrier relationships. A single eSIM profile that follows the device across more than 200 countries removes this layer of administrative overhead. The goal is to provide a single, secure way to stay connected regardless of location, managed through a unified portal.
Satellite constellations have become a viable backhaul for remote telemetry. Early adopters in mining and long-haul trucking already use low-earth-orbit links to monitor asset location and engine health across regions where 4G fails. The projected connection count nears 200 million, but most growth will come from low-power narrowband devices rather than high-throughput links. Manufacturers selling into these verticals are adjusting gateway designs to support dual-mode cellular-plus-satellite routing from the factory floor outward. Automotive is expected to be the largest source of these satellite connections by 2035.
IoT Chip Roadmap Prioritizes RISC-V and Carbon-Aware Design
IoT Analytics released six forecasted semiconductor trends for 2026, highlighting accelerating edge AI integration, chiplet-based architectures, rising adoption of RISC-V cores, carbon-aware chip design, localized production, and security-by-design practices. These trends show a hardware sector that is becoming as flexible as the software it runs.
The move toward RISC-V reflects vendor attempts to reduce dependency on proprietary ISA licensing, particularly as chip shortage cycles continue to pressure supply chains. Several mid-tier IoT siliconmakers have introduced RISC-V-based microcontroller families optimized for low-power sensor fusion workloads. These chips target environmental monitoring, condition-based maintenance, and building automation systems where power envelopes often cap total system draw below one watt. By using an open standard, vendors avoid the licensing fees that can squeeze margins in high-volume commodity markets.
Chiplet architectures allow manufacturers to mix process nodes within a single package, pairing advanced logic dies with older-node analog and RF blocks. That approach cuts per-unit cost while preserving performance for applications requiring simultaneous wireless, signal conditioning, and compute throughput. This modularity is a response to the diverse requirements of the 2026 IoT market, where a single device might need high-speed connectivity and low-power sensing in the same enclosure.
Carbon-aware design remains an emerging theme tied to corporate sustainability commitments. Some fabs now evaluate die placement and thermal profiles against grid emissions intensity to schedule energy-heavy lithography steps during lower-carbon periods. The practice is limited to advanced nodes and does not yet influence commodity sensor controllers, but it signals where the industry expects capital expenditure decisions to land over the next five years. This shift in manufacturing is paired with a growing focus on the energy efficiency of the devices themselves.
Security and Compliance Under the EU Cyber Resilience Act
Security remains a primary concern for the enterprise IoT segment. With the EU Cyber Resilience Act (CRA) 2026 now in effect, vendors must adhere to stricter hardware security requirements. This regulation has forced a cleanup of legacy firmware practices, ensuring that smart home and industrial devices include verifiable update mechanisms and secure-boot processes. The law requires that devices with internet-facing interfaces include vulnerability disclosure timelines and traceable patch cycles.
Compliance is not just a regulatory hurdle but a market advantage. Enterprises are prioritizing vendors who can demonstrate clear audit trails for device updates. The integration of cybersecurity functionalities into industrial edge ecosystems—as seen in recent expansions by companies like Siemens—is now standard operating procedure for any large-scale IoT project. Legacy devices that cannot accept signed firmware updates risk exclusion from European enterprise contracts.
Budget owners should also map software maintenance windows to module EOL dates. On-module ML models need version control just like application code, with rollback images stored on gateways. Teams that treat model weights as release artifacts catch regressions faster and keep field fleets consistent across regions. This structured approach to software lifecycle management is becoming essential as the volume of connected devices grows into the billions.
Summary of the IoT Outlook for 2026
The convergence of standardized protocols and the integration of edge AI means that IoT is finally reaching the maturity required for enterprise-grade deployments. The industry has moved past the experimental phase of individual smart-home gadgets and into an era of integrated, secure, and performant connected infrastructure. The challenge for 2026 is execution: updating legacy networks and ensuring that the new layer of edge intelligence is correctly managed to avoid data sprawl. With the addition of satellite backhaul and on-module AI, the industry is creating a more resilient foundation for the next decade of automation.
Creator: Tech Desk Keywords: IoT, Edge AI, Satellite IoT, 5G RedCap, Telit Cinterion, Cyber Resilience Act, eSIM, RISC-V